1. Timing Performance of Micro-Channel Plates and other detectors with different Front-end Architectures Jean-François Genat CNRS/IN2P3 LPNHE Paris PH-ESE Electronics Seminars CERN Nov. 30th 2011 1

2. This talk will present the major limitations affecting fast timing measurements and the techniques to overcome them. The case of Micro-Channel Plates detectors and accuracies in the picosecond regime will be discussed Fast detectors and signals Timing measurements Threshold techniques Sampling techniques Other techniques Threshold vs Sampling Conclusion 2 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

3. Fast detectors Moving charges: i(t)= n(t) q v(t) i’(t)= q [ n(t) v’(t) + n’(t) v(t)] In order to minimize rise-time, maximize: nmoving charges PMTs, MCPs dv/dt qE/m electric field (in vacuum) dn/dtprimary ionisation multiplication if any: PMTs, MCPs vt. qE /m electric field Fast detectors: - Vacuum devices - Electron multiplication - Low capacitance - High electric fields Bias 3 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

4. Fast detectors Rise-time Timing Regular PMTs 2-5ns 500 ps Silicon APDs, PIN300 ps 50 ps Silicon PMs700 ps 50 ps 3D Silicon 500ps ? Vacuum Multi-anode/mesh PMTs 200ps 50 ps MCP PMTs 100 ps 20-30 ps Streak cameras < 1-10ps Nanosecond: Photo-Multipliers Sub-nanosecond: Silicon Photo-Multipliers Pico-second Micro-Channel Plates, Femto-second Streak camera 4 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

5. Signal development Rise time is I 0 (t) and RC dependent, which set the electronics bandwidth for timing Serial noise proportional to C R small, 50  or less if a common base preamp is used Electronics should not increase C (cables, connectors...) C is the capacitance seen during the rising edge (not a full coaxial cable length) PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures Current signals occur as long as charges move in the detector 5

6. Silicon Photomultipliers 6  Array of avalanche photo diodes: “digital”  photon detection  Array can be 0.5x0.5 up to 5.0x5.0 mm 2  Pixel size can be 10 up to 100  m  All cells connect to a single output  Signal = sum of all cells  Quenching resistor (takes space)  Advantages : Drawbacks:  Good QE (>30%) High dark count  10 6 gain Optical crosstalk  Dead time (Geiger mode) PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

7. Micro-Channel Plate Detectors Pore depth: 1mm 1 st gap pores 2d gap Pore diameter: 3-25  m 200 V 1- 2kV 200 V Anodes (1.6 x 1.6mm 2 pixels) Photo-cathode 10 mm 7 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

8. MCPs signal development Short Transit Time: - Thin photo-cathode gap, - High electric field - Thin MCP: small pore size < 5  m, l < a few 100  m - First strike: cathode on MCP, funnel shaped pore entance Fast pulse: - Thin anode gap, - High electric field MCP signal rising edge: qE = ma l = 1mm, E=100V/mm, tr=250ps PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 8

9. Example: 5 x 5 cm2 10  m-pore Micro-Channel Plate signal Rise time = 300 ps Bandwidth = 0.35/rise-time ~ 1.2 GHz 10% 90% Rise-time - Rise-time - Waveform - Signal to Noise 9 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures Micro-Channel Plate Signals

10. 408nm laser 100 Photo-Electrons 5x5cm 2 Micro Channel Plates from Burle-Photonis Gain: 25  m pores: 0.4mV/PE at 2100 V Rise-time: 650 ps 10  m at 2500V 300 ps The waveform does not depend upon the amplitude Micro-Channel Plate Signals PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 10

11. MCP 25 and 10  m-pore signals spectra 25  m at 2 kV, 50 Photo-electrons 10  m at 2.5 kV “ PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 11

12. MCPs Efficiency and Baseline noise Efficiency plateau and baseline noise (left: 25  m, right 10  m) Plateaux are 250V for both MCPs 10  m MCP shows double and triple after-pulses (not included in the count rates) 25cm 2 MCP ORTEC CFD set at 3PEs 25cm 2 10  m MCP ORTEC CFD set at 3PEs PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 12

13. MCP 5 x 5 cm 2 Burle-Photonis MCP impulse noise rate / cm 2 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 13

14. Fastest Micro-Channel Plate signals 11 mm diameter Micro-Channel Plate signal Signal rise-time/3db bandwidth: 35ps/10 GHz Single Photoelectron Time Transit Spread: 10ps From Photek TTS= 10ps 14 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

15. Outline Fast signals Timing measurements Threshold techniques Sampling techniques Threshold vs Sampling Other techniques Conclusion 15 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

16. Main contributors to timing spreads: Detector: Transit Time fluctuations Noise detector Detector capacitance Rise time fluctuations Amplitude fluctuations Electronics: Noise elec (Amplifiers: thermal, 1/f) ADC: aperture jitter, resolution (ENOB) PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures Timing Measurements 16

17. Timing Measurements Threshold techniques Obtain digital signals from analog detector pulses with analog threshold techniques Digitize using a Time to Digital Coder (TDC) device. Fast, relatively easy to integrate at the Front-end level Sampling techniques Sample and digitize, process in digital Powerful, but needs FPGA computing or integrated in front-ends as mixed-mode ASICs Waveforms digitized using a sampling oscilloscope ASICs can do this today 17 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

18. Electronics gain-bandwidth should match: - Detector sensitivity - Detector rise-time Example: Photek 3.2 mm pore Micro-Channel Plate: Rise-time: 66 ps 3dB bandwidth: 5 GHz Full bandwidth: at least 15 GHz 18 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures Timing Measurements

19. Outline Fast signals Timing measurements Threshold techniques Sampling techniques Threshold vs Sampling Other techniques Conclusion 19 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

20. Threshold Slope dV/dt AA tt noise Effects of noise and on Timing resolution Rise time errors can be partly recovered by peak amplitude measurement or Time Over Threshold 20 Single Threshold PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

21. Single Threshold Leading edge Threshold: Time spread proportional to noise and rise-time 21 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

22. Single Threshold + Peak measurement If the peak amplitude is measured, single threshold can be compensated off-line for rise-time, and even lead to better results than constant fraction 22 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

23. Constant Fraction 1 Threshold at a given fraction of the pulse maximum amplitude OK Leading edge errors Constant fraction Leading edge If rise-time does not depend If pulse shape does not depend upon upon amplitude, leading edge OK amplitude, (most of the cases) use But detector may be under saturation Constant Fraction, analog or digital Leading edge 23 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

24. Constant Fraction Implementation Compare the delayed signal with the attenuated signal (or zero-cross the difference) Need to enable the zero crossing (undefined state before and after operation) A regular discriminator can be used, the delay line is critical (delay, bandwidth), In ICs, it can be implemented with a low pass filter Three parameters: Trigger threshold Delay Fraction Maximize slope at zero-crossing Carefully optimize parameters wrt signals properties Not trivial ! H. Spieler [IEEE NS 29 June 1982 pp1142-1158 ] T.J. Paulus [IEEE NS 32 June 1985 pp 1242-1249] 24 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

25. Improved Threshold Time Pick-off Techniques Extrapolated time Multi-threshold Leading edge errors Leading edge Constant fraction Constant-fraction These techniques can be implemented in digital (FPGAs) after waveform sampling (see slides >30) Threshold at a given fraction of the Several thresholds, reconstruct leading edge and intersect pulse maximum amplitude with time axis PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 25

26. Single Threshold Timing Resolution yield: From slide 18: 26 and Slope is assumed to be constant (not true in most cases !) PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

27. Constant Fraction Timing Resolution 27 Where is the signal slope at fraction f of the amplitude First order expression, assuming constant signal waveform Depends on the detailed implementation PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

28. Multiple Threshold Timing Resolution Since: 28 and PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures Slope is assumed to be constant

29. Discriminator ‘Walk’ Walk: Discriminator delay (walk) depends on slope across threshold Resulting rise-time 2 = ( detector rise-time ) 2 + ( amplifier rise-time ) 2 Use an appropriate technology ( gain x bandwidth ) 29 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

30. Outline Fast signals Timing measurements Threshold techniques Sampling techniques Threshold vs Sampling Other techniques Conclusion 30 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

31. 31 Waveform Sampling 2 12 25 80 128 50 32 … Waveform analysis 31 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

32. 32 Waveform Sampling PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures Preferred to threshold if: - High rate/pile up - Waveform not constant - Pico-second precision needed - No prompt result required Features: - Sample rate - ADC precision - Sampling depth - Triggering - Precise amplitude/charge for free

33. 33 Timing using Waveform Analysis Extract precise time and amplitude from optimal filtering and minimization of  2 evaluated with a fit to a waveform template deduced from averaged measurements Real MCP Laser data Signal Templates Many techniques B. Cleland and E. Stern, BNL 33 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

34. Waveform Sampling Timing Resolution Since 34 and = in case of pure random noise S/N=190 sample rate=10GHz abw=1GHz = 980 fs PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

35. Sampling Electronics for Micro-Channel Plate Detectors Store the full detector information as with a digital oscilloscope: - Detector + electronics noise >> quantization noise (LSB/√12) - Sampling frequency > 2 x full Analog Bandwidth (Shannon-Nyquist) Ideal approach: Digitize on the fly, if the two above conditions can be fulfilled. If not, loss of precision due to A/D conversion and/or loss of timing information 2 GHz Fourier spectrum of a 2”x 2” MCP signal Noise as small as possible Slope as steep as possible 35 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

36. Sampling Electronics for Micro-Channel Plate Detectors A/D converters do not fit the need. State of the art: 8-bit 1GS/s 10-bit 300 MS/s 16-bit 160 MS/s Need at least 5 GS/s sampling rate, 10-12bit There is no ! Fast analog storage and slower digitization, if rate allows, or dead-time acceptable Apply the best timing algorithm suited to the detector, get the charge for free 36 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

37. 37 Fast Sampling Switched Capacitor Array - Sampling frequency - Analog bandwidth - Analog dynamic range - Depth - Readout frequency - Read/Write Switched capacitor array Timing generator Analog input A/D converter 250nm CMOS 6ps rms G. Varner, S. Ritt Chicago-Hawaii: - 130nm CMOS - 15 GS/s sampling rate, 1.5GHz ABW - Picosecond timing resolution foreseen Trigger 37 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 15GS/s 130nm CMOS SCA

38. 38 Signal: 280ps rise-time (Micro-Channel Plate 10  m) Noise: 50% white noise 50% white noise convolved with the MCP response waveform: t exp(-t/  ) Synthesized signals used for simulations * (Matlab) Noise: 50% MCP noise + 50% White noise 38 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures * Available upon request

39. Waveform Sampling: Timing Resolution vs Sampling Rate and ADC Resolution PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures 39

40. Waveform Sampling: Timing Resolution vs Analog Bandwidth and Signal to Noise 40 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

41. 41 Waveform Sampling: Timing resolution vs Sampling rate and Analog Bandwidth (simulation) Timing resolution vs Sampling rate and Analog bandwidth for Sampling rate/Analog bandwidth = 40 41 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

42. Walk: Discriminator delay (walk) depends on slope across threshold Resulting rise-time 2 = ( detector rise-time ) 2 + ( amplifier rise-time ) 2 Use an appropriate technology ( gain x bandwidth ) 42 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures MCP (CFD measured)

43. CMOS ICs Bandwidths CMOS bandwidths from 90 to 45 nm technology nodes (ITRS 2005) From 12 to 62 GHz 43 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

44. 44 Waveform Sampling benefits Pulse sampling and waveform analysis for 2D delay line readout Pico-second timing with fast detectors Resolve pile-up, mishappened pulses Timing along the detector, a few ps obtained (< 500  m, 18 Photo-electrons) Centroids perpendicular Large area detectors can be read in series 44 Delay lines as anodes PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

45. 45 Position resolution with 5x5 cm 2 MCPs read by timing Measured position resolution (velocity=8.25ps/mm) : 50PEs 4.26ps 213  m 5x5cm 2 Micro Channel Plates 158PEs 1.95ps 97  m 45 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

46. Outline Fast signals Timing measurements Threshold techniques Sampling techniques Threshold vs Sampling Other techniques Conclusion 46 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

47. Timing techniques compared * (Matlab) Monte-Carlo for time resolution vs number of photo-electrons using: - Single threshold - Multiple threshold - Constant fraction - Waveform Sampling Input signals 47 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures zoom Noise is 50% white noise 50% MCP shaped noise

48. 48 Threshold vs Sampling Hardware Architectures Present trends: Move Signal Processing to Front-end and Amplifier to Detector PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

49. Outline Fast signals Timing measurements Threshold techniques Sampling techniques Threshold vs Sampling Other techniques Conclusion 49 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

50. Zero crossing Use zero-crossing of signal derivative Detects signal’s maximum Might not be what is needed ! Signal delay Reject noise from signal derivative 50 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

51. Double Threshold High threshold to trigger Low threshold to time Low thresh Hi thresh Delay Signal delay Avoids noise on low threshold Decision on very first signal rise 51 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

52. Outline Fast signals Timing measurements Threshold techniques Sampling techniques Threshold vs Sampling Other techniques Conclusion 52 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

53. 53 Timing pick-off requires dedicated analog front-end electronics matched with the detector signal characteristics in terms of noise and bandwidth. The key number is the slope to noise ratio Many techniques exist matched to different detector and environment conditions Choice is also dictated by environment constraints such as event rate, number of channels, availability of digital signal processing and ASIC design means, and cost 53 Conclusion PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

54. - J.L. Wiza. Micro-channel Plate Detectors. Nucl. Instr. Meth. 162 (1979) 587-601. - K. Inami, N. Kishimoto, Y. Enari, M. Nagamine, and T. Ohshima. Timing properties of MCP-PMT. Nucl. Instr. Meth. A560 (2006) 303-308., K. Inami. Timing properties of MCP-PMTs. Proceedings of Science. International Workshop on new Photon-Detectors, June 27-29 (2007). Kobe University, Japan. - J. Va’vra, J. Benitez, J. Coleman, D. W. G. Leith, G. Mazaher, B. Ratcliff and J. Schwiening. A 30 ps Timing Resolution for Single Photons with Multi-pixel Burle MCP-PMT. Nucl. Instr. Meth. A572 (2007) 459-462. - H. Kim et al. Electronics Developments for Fast Timing PET Detectors. Symposium on Radiation and Measurements Applications. June 2-5 (2008), Berkeley CA, USA. An extensive list of references on timing measurements can be found in: A.Mantyniemi, MS Thesis, Univ. of Oulu, 2004; ISBN 951-42-7460-I; ISBN 951-42-7460-X; - S. Cova et al. Constant Fraction Circuits for Picosecond Photon Timing with Micro-channel Plate Photomultipliers. Review of Scientific Instruments, 64-1 (1993) 118-124. - D. Breton, E. Delagnes, J. Maalmi, K. Nishimura, L.L Ruckman, G.S. Varner, Va’Vra. High Resolution Photon Timing with MCP-PMT. Proceedings of NSS-MIC 2010 Conference, Knoxville TN, USA - S. Ritt. Design and Performance of the 5 GHz Waveform Digitizer Chip DRS3. Nuclear Instruments and Methods, (2007). - W.E. Cleland, E.G. Stern. Signal Processing Considerations for Liquid Ionization Calorimeters in a High Rate Environment, Nuclear Instruments and Methods A338 pp 467-497 - G. Varner, L.L. Ruckman, A. Wong. The First version Buffered Large Analog Bandwidth (BLAB1) ASIC for high Luminosity Colliders and Extensive Radio Neutrino Detectors. Nucl. Inst. Meth. A591 (2008) 534. - G. Bondarenko, B. Dolgoshein et al. Limited Geiger Mode Silicon Photodiodes with very high Gain. Nuclear Physics B, 61B (1998) 347-352. - J-F Genat, G. Varner, F. Tang and H.J. Frisch. Signal Processing for Pico-second Resolution Timing Measurements. Nuclear Instruments and Methods, (2009). - J. Christiansen. An Integrated CMOS 0.15 ns Digital. Timing Generator for TDC's and Clock Distribution. Systems, IEEE Trans. Nucl. Sci., Vol. 42, No4 (1995), p. 753 54 References PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures

55. 55 Thanks ! 55 PH-ESE Electronics Seminars, Nov. 30th 2011, CERN Fast Timing and Front-end Architectures